README
----------


A clear application of computing to traffic is real-time optimization of traffic signal optimization. 
However, this kind of problem requires the evaluation of the quality of various potential solutions. 
While this is not difficult, it is currently extremely slow rendering real time solution infeasible. 
In this project we propose to replace current microsimulation solutions with a meso simulator using 
machine learning to learn the arrival time function for traffic parameters on a street segment by 
street segment characteristics. We then run a priority queue-based simulator that estimates the 
arrival time of each car at the next decision point for that car. We tested our simulator against 
data collected from the widely used VISSIM micro simulator. A big advantage of this solution is the 
customization of the arrival time predictor to specific street segments and other parameters such as
 time of day or weather. This would replace the one-size-fits-all formulas used by most simulators.